Pulse Coupled Neural Networks and Image Morphology for Mammogram Preprocessing

Given that over 230,000 women in the United States alone will contract breast cancer, resulting in over 39,000 deaths and that there will be an estimated 458000 such deaths worldwide, the early detection and management of breast cancer is a significant problem. Currently, mammography provides the dominant front-line screening procedure. To assist in the interpretation of mammograms, a variety of computer aided diagnostic algorithms have been developed. A critical step in most of these algorithms is to remove image artifacts and isolate the breast from the mammogram background. This study explores the use of a biologically inspired model, the Pulse Coupled Neural Network, to form candidate image segments that, when combined with standard image morphology operators, can be used to remove image acquisition artifacts and isolate the breast profile in the mammogram.